臺灣博碩士論文加值系統

目的 : 本研究針對乳牛產乳量下降之乳牛生理、環境等因素進行三種不同預測模型分析及比較，並透過多種預測效力檢定評估預測模型之模型效力，以提供產乳量下將之風險預測模型，最後將研究成果回饋至自動化熱影像監控平台，作為牧場乳牛照護監測之產乳量下降警示依據。

材料與方法 : 本研究蒐集320筆乳牛個體資料 (體重、眼窩體溫、年齡 (月)、泌乳天數、胎次)、外在環境因子 (溫度、濕度、熱緊迫溫溼度指數 (temperature and humidity index, THI)) 及產乳量資訊，使用最小絕對壓縮挑選運算子 (Least absolute shrinkage and selection operator, LASSO) 及隨機森林演算法進行因子重要程度分析與排序，並針對重要因子結合向前選取法建立單/多變數邏輯斯回歸模型及隨機森林預測模型，並使用受測者操作特徵曲線下面積 (Area under the receiver operating characteristic curve, AUC)、準確率、平均絕對誤差 (Mean absolute error, MAE)、均方誤差 (Mean squared error, MSE)、均方根誤差 (Root mean squared error, RMSE) 、R-square、陽性預測值 (Positive predictive value, PPV) 與陰性預測值 (Negative predictive value, NPV)評估模型之預測效力。

結果 : 本研究以LASSO挑選之最重要因子進行單變數邏輯斯模型建立，模型參數之20% 發生產乳量下降之眼窩體溫TT20 = 37.2°C (36.4 ~ 38.0)、50% 發生產乳量下降之眼窩體溫TT50 = 39.9 °C (39.1 ~ 40.7) 該模型預測效力1) AUC為0.687、2) 準確率為0.703、3) MAE為 128.347、4) MSE為 64.058、5) RMSE為0.447及6) R-squared為0.096；在多變數研究之向前選取法邏輯斯回歸模型所納入的因子為 1) 眼窩體溫、2) 體重及 3) 溫度，該模型預測效力1) AUC為0.741、2) 準確率為0.716、3) MAE為 116.986、4) MSE為 58.700、5) RMSE為0.428及6) R-squared為0.172；隨機森林模型所納入的因子為 1) 眼窩體溫、2) 體重、3) 泌乳天數、4) 年齡 (月) 及 5) 濕度，該模型預測效力 1) AUC為0.891、2) 準確率為0.819、3) MAE為 104.194、4) MSE為 44.382、5) RMSE為0.372及6) R-squared為0.374。

結論 : 本研究針對乳牛眼窩溫度建立單變數邏輯斯預測模型並定義50%機率發生產乳量下降之眼窩體溫為39.9 °C，同時也針對場域及乳牛個體因子建立多變數預測模型，其中以隨機森林預測模型具有最佳的預測效力，主要挑選出五項重要因子分別為: 1) 眼窩體溫、2) 體重、3) 泌乳天數、4) 年齡 (月) 及 5) 濕度。該健康資訊可給予畜牧養殖場做為參考指標以改良照護品質，進而可提升乳牛孕育胎數增加牛乳產量，本研究之貢獻可增加畜牧業之實體經濟效益。

Purpose : The research, focused on physiological and environmental factors decreasing the amount of milk supply from dairy cattles, mainly analysed and compared three prediction models. It provides risk prediction models about the decreasing amount of milk supply through several tests evaluating the effects of prediction models. Returning the research results to the automatic thermal image monitoring platform as the data of decreasing milk supply in dairy cattles caring.
Materials and methods : The research compiled 320 piece of data, including individual statistics of dairy cattles (Weight, periorbital temperature, age(month), lactation days and birth parities), outside environmental factors (temperature, humidity, Temperature and Humidity Index (THI) and milk supply information. Analyse and sort importance of factors through Least Absolute Shrinkage and Selection Operator (LASSO) and random decision forests. Set single or several variables logistic regression models through forward selection to focus on essential factors. Use Area under the receiver operating characteristic Curve (AUC), accuracy, Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Positive predictive value (PPV), Negative predictive value (NPV) and predictable effects of R-square evaluation models.
Results : The research was set by using the chosen important factors from LASSO to run single or several variables logistic regression models. 20% of parameters happened under the periorbital temperature 37.2°C of decreasing milk supply (TT20 = 37.2°C (36.4 ~ 38.0)), 50% of parameters happened under the periorbital temperature 39.9 °C of decreasing milk supply ( TT50 = 39.9 °C (39.1 ~ 40.7)). The predictable effects: 1) AUC is 0.687, 2) accuracy is 0.703, 3) MAE is 128.347, 4) MSE is 64.058, 5) RMSE is 0.447and 6) R-squared is 0.096. Several variables of logistic regression models through forward selection include 1) periorbital temperature, 2) weight and 3) temperature. The predictable effects of the model include 1) AUC is 0.741, 2) accuracy is 0.716、3) MAE is 116.986, 4) MSE is 58.700, 5) RMSE is 0.428 and 6) R-squared is 0.172. Random decision forests include 1) periorbital temperature, 2) weight, 3) lactation days, 4) age(month) and 5) humidity. The predictable effects of the model include 1) AUC is 0.891, 2) accuracy is 0.819, 3) MAE is 104.194, 4) MSE is 44.382, 5) RMSE is 0.372 and 6) R-squared is 0.374.
Conclusions : The research set single variables logistic regression models for periorbital temperature of milk cattles and define that decreasing milk supply happened under the periorbital temperature is 39.9 °C. At the same time, set several variables logistic regression models for the field area and individual factors of milk cattles. Receive the best predictable effects from random decision forests, which the five main essential factors are1) periorbital temperature, 2) weight, 3) lactation days, 4) age(month) and 5) humidity. The health information can be delivered to livestock farms as reference standards to improve caring quality, which can increase birth parities to increase milk supply. The research can exactly increase the economic benefits.

摘要 i
Abstract iii
致謝 vi
目錄 vii
表目錄 ix
圖目錄 x
符號縮寫 xi
第一章 緒論 1
1.1 動機 1
1.2 目的 2
1.3 相關文獻探討 3
第二章 材料與方法 5
2.1 前言 5
2.2 乳牛資料 8
2.2.1 硬體設備架設與數據 9
2.3 因子萃取及模型建立 11
2.3.1 最小絕對壓縮挑選運算子 (LASSO) 11
2.3.2 單變數邏輯斯回歸模型 12
2.3.3 多變數邏輯斯回歸模型 12
2.3.4 隨機森林模型 13
2.4 模型預測效力評估 15
2.4.1 AUC 15
2.4.2 準確率 15
2.4.3 平均絕對誤差 (Mean Absolute Error, MAE) 15
2.4.4 均方誤差 (Mean Squared Error, MSE) 16
2.4.5 均方根誤差 (Root Mean Squared Error, RMSE) 16
2.4.6 R-squared 16
2.4.7 陽性預測值(PPV)與陰性預測值(NPV) 17
第三章 結果 18
3.1 前言 18
3.2 LASSO因子重要程度分析 19
3.3 單變數邏輯斯回歸模型 21
3.4 向前選取法多變數邏輯斯回歸模型 22
3.5 隨機森林回歸模型 23
3.6 模型預測效力評估 26
3.7 成果回歸 27
第四章 討論 28
4.1 前言 28
4.2 乳牛體溫量測 28
4.3 體重參數 31
4.4 環境溫度濕度THI 32
4.5 泌乳天數 33
第五章 結論 35
參考文獻 36

[1]J. W. West, "Effects of heat-stress on production in dairy cattle," Journal of dairy science, vol. 86, pp. 2131-2144, 2003.
[2]I. García-Ispierto, F. López-Gatius, G. Bech-Sabat, P. Santolaria, J. Yániz, C. Nogareda, et al., "Climate factors affecting conception rate of high producing dairy cows in northeastern Spain," Theriogenology, vol. 67, pp. 1379-1385, 2007.
[3]A. Lees, V. Seijan, A. Wallage, C. Steel, T. Mader, J. Lees, et al., "The impact of heat load on cattle: Review. Animals. 9: 322. ISSN: 2076-2615," ed, 2019.
[4]J. Pryce, R. Veerkamp, R. Thompson, W. Hill, and G. Simm, "Genetic aspects of common health disorders and measures of fertility in Holstein Friesian dairy cattle," Animal Science, vol. 65, pp. 353-360, 1997.
[5]J. Windig, M. Calus, and R. Veerkamp, "Influence of herd environment on health and fertility and their relationship with milk production," Journal of dairy science, vol. 88, pp. 335-347, 2005.
[6]O. Burfeind, M. Von Keyserlingk, D. Weary, D. Veira, and W. Heuwieser, "Repeatability of measures of rectal temperature in dairy cows," Journal of dairy science, vol. 93, pp. 624-627, 2010.
[7]B. I. Smith and C. A. Risco, "Management of periparturient disorders in dairy cattle," Veterinary Clinics: Food Animal Practice, vol. 21, pp. 503-521, 2005.
[8]N. Burdick, J. Carroll, J. Dailey, R. Randel, S. Falkenberg, and T. Schmidt, "Development of a self-contained, indwelling vaginal temperature probe for use in cattle research," Journal of Thermal Biology, vol. 37, pp. 339-343, 2012.
[9]台灣動物社會研究會. (2019). 台灣乳牛飼養五大痛苦指標. Available: https://www.east.org.tw/action/8355
[10]G. Esposito, P. C. Irons, E. C. Webb, and A. Chapwanya, "Interactions between negative energy balance, metabolic diseases, uterine health and immune response in transition dairy cows," Animal reproduction science, vol. 144, pp. 60-71, 2014.
[11]J. R. Roche, N. C. Friggens, J. K. Kay, M. W. Fisher, K. J. Stafford, and D. P. Berry, "Invited review: Body condition score and its association with dairy cow productivity, health, and welfare," Journal of dairy science, vol. 92, pp. 5769-5801, 2009.
[12]沈昱名, 詹昆衛, 楊瑋誠, 蘇耀期, 余章游, 王建雄, et al., "綜論各國之乳牛體態評分標準與常見疾病之相關性," 臺灣獸醫學雜誌, vol. 39, pp. 135-149, 2013.
[13]W. Rauw, E. Kanis, E. Noordhuizen-Stassen, and F. Grommers, "Undesirable side effects of selection for high production efficiency in farm animals: a review," Livestock production science, vol. 56, pp. 15-33, 1998.
[14]卓盈旻 and 盧孟明, "IPCC 第四期評估報告氣候模式對台灣氣候變遷模擬之分析," 2010.
[15]E. Gernand, S. König, and C. Kipp, "Influence of on-farm measurements for heat stress indicators on dairy cow productivity, female fertility, and health," Journal of dairy science, vol. 102, pp. 6660-6671, 2019.
[16]林育安. (2014). 高溫環境對乳牛生理和泌乳影響，103年. Available: https://www.coa.gov.tw/ws.php?id=2501744
[17]陳志毅, "DHI 牛群生產年限風險因子之探討
Study on risking factors of production life for DHI," 生產年限；疾病與淘汰；DHI prosuction life；disease and culling；Dairy herd improvement 2012.
[18]陳志峰. (2019). 從擠乳機器人看農業缺工——引入農業外勞不是唯一選項. Available: https://www.agriharvest.tw/archives/19575
[19]R. Tibshirani, "Regression shrinkage and selection via the lasso," Journal of the Royal Statistical Society. Series B (Methodological), pp. 267-288, 1996.
[20]R. Tibshirani, M. Saunders, S. Rosset, J. Zhu, and K. Knight, "Sparsity and smoothness via the fused lasso," Journal of the Royal Statistical Society: Series B (Statistical Methodology), vol. 67, pp. 91-108, 2005.
[21]H. Zou and T. Hastie, "Regularization and variable selection via the elastic net," Journal of the Royal Statistical Society: Series B (Statistical Methodology), vol. 67, pp. 301-320, 2005.
[22]E. Scornet, G. Biau, and J.-P. Vert, "Consistency of random forests," The Annals of Statistics, vol. 43, pp. 1716-1741, 2015.
[23]Q. Hu, X.-S. Si, Q.-H. Zhang, and A.-S. Qin, "A rotating machinery fault diagnosis method based on multi-scale dimensionless indicators and random forests," Mechanical Systems and Signal Processing, vol. 139, p. 106609, 2020.
[24]B. T. Pham, K. Khosravi, and I. Prakash, "Application and comparison of decision tree-based machine learning methods in landside susceptibility assessment at Pauri Garhwal Area, Uttarakhand, India," Environmental Processes, vol. 4, pp. 711-730, 2017.
[25]S. Talukdar, K. U. Eibek, S. Akhter, S. Ziaul, A. R. M. T. Islam, and J. Mallick, "Modeling fragmentation probability of land-use and land-cover using the bagging, random forest and random subspace in the Teesta River Basin, Bangladesh," Ecological Indicators, vol. 126, p. 107612, 2021.
[26]T. Shaikhina, D. Lowe, S. Daga, D. Briggs, R. Higgins, and N. Khovanova, "Decision tree and random forest models for outcome prediction in antibody incompatible kidney transplantation," Biomedical Signal Processing and Control, vol. 52, pp. 456-462, 2019.
[27]R. J. Lewis, "An introduction to classification and regression tree (CART) analysis," in Annual meeting of the society for academic emergency medicine in San Francisco, California, 2000.
[28]N. Farnaaz and M. Jabbar, "Random forest modeling for network intrusion detection system," Procedia Computer Science, vol. 89, pp. 213-217, 2016.
[29]K. G. Moons, A. P. Kengne, M. Woodward, P. Royston, Y. Vergouwe, D. G. Altman, et al., "Risk prediction models: I. Development, internal validation, and assessing the incremental value of a new (bio) marker," Heart, vol. 98, pp. 683-690, 2012.
[30]V. Kamble and S. Deshmukh, "Comparision Between Accuracy and MSE, RMSE by Using Proposed Method with Imputation Technique," Oriental Journal of Computer Science and Technology, vol. 10, pp. 773-779, 2019.
[31]D. Daneshvar and A. Behnood, "Estimation of the dynamic modulus of asphalt concretes using random forests algorithm," International Journal of Pavement Engineering, pp. 1-11, 2020.
[32]D. C. Montgomery, E. A. Peck, and G. G. Vining, Introduction to linear regression analysis: John Wiley & Sons, 2021.
[33]J. M. Naylor, R. M. Streeter, and P. Torgerson, "Factors affecting rectal temperature measurement using commonly available digital thermometers," Research in Veterinary Science, vol. 92, pp. 121-123, 2012.
[34]G. Hoffmann, M. Schmidt, C. Ammon, S. Rose-Meierhöfer, O. Burfeind, W. Heuwieser, et al., "Monitoring the body temperature of cows and calves using video recordings from an infrared thermography camera," Veterinary research communications, vol. 37, pp. 91-99, 2013.
[35]L. Vickers, O. Burfeind, M. Von Keyserlingk, D. Veira, D. Weary, and W. Heuwieser, "Comparison of rectal and vaginal temperatures in lactating dairy cows," Journal of dairy science, vol. 93, pp. 5246-5251, 2010.
[36]A. Schaefer, N. Cook, C. Bench, J. Chabot, J. Colyn, T. Liu, et al., "The non-invasive and automated detection of bovine respiratory disease onset in receiver calves using infrared thermography," Research in veterinary science, vol. 93, pp. 928-935, 2012.
[37]B. Horan, P. Dillon, P. Faverdin, L. Delaby, F. Buckley, and M. Rath, "The interaction of strain of Holstein-Friesian cows and pasture-based feed systems on milk yield, body weight, and body condition score," Journal of dairy science, vol. 88, pp. 1231-1243, 2005.
[38]J. Roche, J. Lee, K. Macdonald, and D. Berry, "Relationships among body condition score, body weight, and milk production variables in pasture-based dairy cows," Journal of Dairy Science, vol. 90, pp. 3802-3815, 2007.
[39]R. Bouraoui, M. Lahmar, A. Majdoub, and R. Belyea, "The relationship of temperature-humidity index with milk production of dairy cows in a Mediterranean climate," Animal Research, vol. 51, pp. 479-491, 2002.
[40]R. Zimbelman, R. Rhoads, M. Rhoads, G. Duff, L. Baumgard, and R. Collier, "A re-evaluation of the impact of temperature humidity index (THI) and black globe humidity index (BGHI) on milk production in high producing dairy cows," in Proceedings of the Southwest Nutrition Conference (ed. RJ Collier), 2009, pp. 158-169.
[41]A. F. Groen, T. Steine, J.-J. Colleau, J. Pedersen, J. Pribyl, and N. Reinsch, "Economic values in dairy cattle breeding, with special reference to functional traits. Report of an EAAP-working group," Livestock Production Science, vol. 49, pp. 1-21, 1997.
[42]J. Dekkers, J. Ten Hag, and A. Weersink, "Economic aspects of persistency of lactation in dairy cattle," Livestock Production Science, vol. 53, pp. 237-252, 1998.
[43]J. Sölkner and W. Fuchs, "A comparison of different measures of persistency with special respect to variation of test-day milk yields," Livestock Production Science, vol. 16, pp. 305-319, 1987.